High Spectral Resolution Raman Measurements Using Light-Emitting Diode as Excitation Based on Weighted Spectral Reconstruction Method

Shuo Chen,Jing Liu,Wenbin Xu,Gang Hou,Xiaoyu Cui,Liping Xie,Lingmin Kong,Mixing Liu

doi:10.1109/access.2019.2941262

Abstract

Raman spectroscopy is a spectroscopic technique that can provide rich biochemical information about the sample through the inelastic scattering between monochromatic light and molecular vibrations, thus a narrow bandwidth and costly laser is typically required. Although the replacement of laser to light-emitting diode (LED) can reduce the cost significantly, spectral resolution degrades seriously and a large amount of Raman fingerprint information is lost due to the overlaps of Raman peaks. In this study, a weighted spectral reconstruction method was developed to retrieve high quality Raman spectrum from the Raman measurements using LED as the excitation, in which the LED based Raman measurements were numerically synthesized from the Raman spectra acquired by the conventional laser-based Raman spectroscopy systems. The proposed weighted spectral reconstruction method was numerically test on the synthesized LED based Raman measurements from 25 agar phantoms, 50 blood serum samples, and 56 cell samples, respectively. According to the results, the restored Raman spectra are in excellent agreement with the conventional Raman spectra, and the differences are always less than 2% in all test samples. Therefore, Raman measurements using LED as excitation combined with weighted spectral reconstruction method demonstrate significant potential in accurate and high spectral resolution Raman measurements, which promotes a cost-effective Raman setup while preserve the most important spectral and biochemical information.

Highlights

Raman spectroscopy is a non-invasive vibrational spectroscopic technique and can provide quantitative fingerprint information about molecular vibrations, which relies on the inelastic scattering (i.e. Raman scattering) between the monochromatic photons and molecules [1]
It is expected that the recovery of high spectral resolution Raman spectra from Light-emitting diode (LED) based Raman should be more difficult with the increment of LED bandwidth and Raman spectral complexity
Since the LED has to be existed in real-world practical applications other than those numerically simulated non-negative principal components (PCs) based filters, the choice of LED bandwidth should be based on the compromise between the availability of the LED bandwidth and spectral reconstruction accuracy

Summary

Introduction

Raman spectroscopy is a non-invasive vibrational spectroscopic technique and can provide quantitative fingerprint information about molecular vibrations, which relies on the inelastic scattering (i.e. Raman scattering) between the monochromatic photons and molecules [1]. In the training stage for weight calculation in this study, Wiener matrix W is created to learn supervised from the training dataset according to (2), in which the model between the conventional high spectral resolution Raman spectra r and the corresponding LED based narrow-band measurements m in the calibration dataset can be built.

Results

Conclusion